Post-Exercise Cardiac Recovery as a Multiscale Dynamical Process

Nonlinear dynamics and chaos-based signal analysis provide a powerful framework for probing physiological regulation beyond conventional linear descriptors. In particular, post-exercise recovery represents a controlled, reproducible dynamical transition through which the adaptability and coupling of cardiovascular and autonomic subsystems can be examined. We investigate recovery dynamics using multifractal spectra (MFS) of simultaneously recorded multimodal cardiovascular signals—electrocardiography (ECG), photoplethysmography (PPG), seismocardiography (SCG), and phonocardiography (PCG)—as illustrated in Fig. 1. Thirty-second recordings from 28 healthy volunteers, acquired at rest and during post-exercise recovery from the SensSmartTech database, were analyzed. MFS features were extracted to characterize scale-dependent fluctuations arising from nonlinear interactions across modalities. To assess the discriminative power of these features, 5 supervised learning algorithms—Logistic Regression, Support Vector Machine with RBF kernel, kNearest Neighbours, Decision Tree, and Random Forest—were employed to classify recovery states. Two complementary feature sets were evaluated with a target to classify subjects into three groups defined by the speed of recovery: Fchange, capturing long-term dynamical adaptation between resting and post-exercise states, and Fhalf, quantifying recovery at a standardized half-recovery time. Our results demonstrate a reliable classification using a compact set of MFS features, even with limited data duration, and a clear benefit from inclusion of nonlinear interactions. They indicates that the postexercise recovery is governed by coupled, multiscale processes spanning multiple physiological subsystems. The integration of multifractal analysis with multimodal sensing thus provides robust, physically interpretable markers of reco